Sorting a mixed inbox into schemas
A modelled scenario: an unsorted set of invoices, receipts and contracts is classified into schemas and grouped into smart folders.
The manual process
- Invoices, receipts, contracts and other files arrive together in a shared inbox or drive.
- Someone sorts them by hand and keys the main details into a spreadsheet or accounting tool.
- Duplicates and unusual documents are noticed only when something goes wrong.
The pipeline
-
Upload the mix
Files are uploaded in the portal or sent via API and webhooks, with no sorting beforehand.
-
Classify into schemas
The general business documents pack provides 14 ready schemas, and each document is matched to one.
-
Extract and index
Fields are extracted with confidence scores. The pack also provides 8 indexes for values such as issuer, total and currency.
-
Handle the unknown
A document that matches no schema is flagged. A reviewer can assign a schema or generate a new one, and reprocess it.
-
Group and ask
The pack provides 6 smart folders and 12 saved questions, such as finding duplicate reference numbers.
The result
- Documents land in the right schema without manual sorting.
- Related documents are grouped in smart folders.
- Unrecognised documents are surfaced for a decision.
What the reviewer sees
A flagged list for unknown or low-confidence documents, with the extracted fields beside each page, and folders that fill as documents are approved.
What this scenario does not claim
This is an illustration, not a customer case study, with no volumes, savings or accuracy figures. The pack counts describe what ships in the general pack, not an outcome. AI output can be wrong, so review remains part of the flow.